Mantle API presentation by AMD, DICE and Oxide - AMD Summit 2013

[Fourstaff]

From what I gather it's not. Mantle is it's own thing, like Driect X. Nvidia gets their architecture to work with Direct X, so does AMD. Nvidia just needs to make their architecture work with Mantle.

...unless I'm totally wrong, which is always a possibility.

So if we get open source Mantle, OpenCL will be consigned to history?

 

[bwat47]

So if we get open source Mantle, OpenCL will be consigned to history?

No, OpenCL and Mantle are used for different things. OpenCL is more like an open source version of Nvidia's CUDA, its used for GPU Compute stuff. Mantle is more like OpenGL or DIrectX (but lower-level than directx).

 

[bpgt64]

I don't completely understand the difference between OpenGL, OpenCL, DirectX, Mantle, etc.... It makes me wonder how this shifting of rendering methods will be effected if a substantial amount of gamers go to Linux with SteamOS on top of Android's OS exploding.

 

[CD2 Solutions]

I don't completely understand the difference between OpenGL, OpenCL, DirectX, Mantle, etc.... It makes me wonder how this shifting of rendering methods will be effected if a substantial amount of gamers go to Linux with SteamOS on top of Android's OS exploding.

with any luck, mantle will move to the ARM powered sectors too.

openGL directx and mantle are all used to make graphics engines and games easier to make for a wide variety of hardware. without them, a developer would need to make a different version of the game for EVERY graphics chip out there, something that is frankly ludicrous. by using a graphics api, it gives the developer a wider coverage, only having to adjust a little for different architectures (amd cards from the 7xxx series for example share the same basic architecture).

openCL does a similar thing, but its for compute tasks like video conversions.
mantle is open as well, so a linux version will just require porting. how long that will take i dont know, but with a company like valve behind it.........

PS: i hope you werent being sarcastic lol. either way, somebody who doesnt know will read this.

 

[bpgt64]

with any luck, mantle will move to the ARM powered sectors too.

openGL directx and mantle are all used to make graphics engines and games easier to make for a wide variety of hardware. without them, a developer would need to make a different version of the game for EVERY graphics chip out there, something that is frankly ludicrous. by using a graphics api, it gives the developer a wider coverage, only having to adjust a little for different architectures (amd cards from the 7xxx series for example share the same basic architecture).

openCL does a similar thing, but its for compute tasks like video conversions.
mantle is open as well, so a linux version will just require porting. how long that will take i dont know, but with a company like valve behind it.........

PS: i hope you werent being sarcastic lol. either way, somebody who doesnt know will read this.

I wasn't being sarcastic at all. My experience with Graphics APi's has been, well this runs on Direct X 9 so it generally looks so-good. Where as some of the games i have seen on DX11 look pretty amazing. All I know is what I experience in different games(varying based on developer/execution).

Thank you for the explanation.

 

[FordGT90Concept]

I think the best synonym for Mantle is x86. I believe Mantle exposes a set of instructions software can use directly in the GPU instead of going through layers of APIs. As long as the GPU supports those instructions (I highly doubt NVIDIA and Intel do but they can add it in future silicon), Mantle will work.

 

[TheoneandonlyMrK]

Well, SteamBox idea isn't stupid you know. I have my Lian Li placed in a living room right now and it looks like it belongs there from day one. It doesn't even look much as a PC in a smaller black ion platted case. It looks more like a HiFi component.

So bottom line, for SteamBox it's all about what they pack in it. For countless indie games on Steam, you really don't need much of a hardware. Any half decent AMD Fusion system could do the job at 1080p. For high end games, it really depends.

I've also tried Big Screen Steam and it looks rather nice. So, clearly, there is a potential.

no one really sits in front of a 42-50" living room telly checking facebook and playing peggle(thats why we sit with pads/phones),, they have consoles and its consoles that they (steambox's)have to contend with in performance terms,at least imho and as good as open Gl is I think Steam need to climb aboard the HSA / Mantle train ,to optimise there bang for buck with Amd based steambox's.

 

[bpgt64]

no one really sits in front of a 42-50" living room telly checking facebook and playing peggle(thats why we sit with pads/phones),, they have consoles and its consoles that they (steambox's)have to contend with in performance terms,at least imho and as good as open Gl is I think Steam need to climb aboard the HSA / Mantle train ,to optimise there bang for buck with Amd based steambox's.

I think Valve just needs get something on Debian based Linux and let developers decide which makes better use of end-hardware, and in turn let development of games drive which gets used more.

 

[RejZoR]

no one really sits in front of a 42-50" living room telly checking facebook and playing peggle(thats why we sit with pads/phones),, they have consoles and its consoles that they (steambox's)have to contend with in performance terms,at least imho and as good as open Gl is I think Steam need to climb aboard the HSA / Mantle train ,to optimise there bang for buck with Amd based steambox's.

I do? I have my BlackBox rigged to 42 inch LCD and i do everything on it right now. Did increase the DPI to see smalls tuff easier...

 

[RoutedScripter]

Open standard doesn't mean anything if GCN is the only architecture able to support Mantle.

From what I gather it's not. Mantle is it's own thing, like Driect X. Nvidia gets their architecture to work with Direct X, so does AMD. Nvidia just needs to make their architecture work with Mantle.

...unless I'm totally wrong, which is always a possibility.

You are correct, the code that is there currently only supports GCN, all you have to do is to include code for other architectures, it has nothing to do with GCN code already sitting beside it, Nvidia would have to include and develop code from scratch, AMD will in any case have at least 2 years of advantage, possibly more if Nvidia is arrogant enough to not take this seriously, which looking at it's history, probably will happen.

I don't believe Khronos is the best group to oversee something like Mantle considering they have done a miserable job keeping OpenGL updated and running properly...

I fully agree with that. That was a terrible thing to ask. Since the Mantle presentations i've been on several public chats and extensively discussed Mantle with OGL community people, even they aren't happy with Khronos at all.

So if we get open source Mantle, OpenCL will be consigned to history?

Depends to what low level the GPU gets unlocked, let's not forget this is not 100% like on consoles, it's still some abstraction going on, but I do think that these tasks can be done manutally via Mantle anyways, and if it's lower level it might be even better than with OpenCL which is supposably higher-level, however Carmack pointed out how Nvidia is quick to include new OGL/OCL extensions for devs, I don't think I am suitable to discuss this more than this, I do not have full understanding.

It makes me wonder how this shifting of rendering methods will be effected if a substantial amount of gamers go to Linux with SteamOS on top of Android's OS exploding.

That is not certain, SteamOS is now more tied to Nvidia's OGL polish, the efficiency and performance of Linux OS alone probably will not be enough to beat Mantle on windows. The most powerful available (not possible, because everything is possible if you want to make it, talking about in present time) combination in the industry will is Mantle and Linux.

Android OS is irrelevant.

 

[bpgt64]

The only way it seems to tell which approach is better, is to watch AMD and see if they can reclaim the performance crown with Mantle and AMD CPUs while nvidia sticks with intel CPUs and uses OpenGL/Linux approach.

 

[TheMailMan78]

Its all smoke and mirrors until W1zz reviews. Also did I just read Android OS is irrelevant? I sure hope your kidding.

 

[Frag_Maniac]

The only way it seems to tell which approach is better, is to watch AMD and see if they can reclaim the performance crown with Mantle and AMD CPUs while nvidia sticks with intel CPUs and uses OpenGL/Linux approach.

Mantle's development was primarily GPU oriented. It only frees up CPU cycles in the process due to being less dependent on the CPU by extension of more GPU efficiency. There should be no change in how AMD CPUs fare against Intel's under Mantle.

 

[TheoneandonlyMrK]

I think Valve just needs get something on Debian based Linux and let developers decide which makes better use of end-hardware, and in turn let development of games drive which gets used more.

you know whilst i implied it as a recomendation my opinion is that valve Are working on utilising mantle and hsa

Mantle's development was primarily GPU oriented. It only frees up CPU cycles in the process due to being less dependent on the CPU by extension of more GPU efficiency. There should be no change in how AMD CPUs fare against Intel's under Mantle.

Whilst initially id expect what your saying to hold out i dont think that will remain true with time ,who after all knows Amd's 64bit arch and instruction set best , plus Mantle is like the very tip of the corner(just 1) of a pyramid Amd are trying to construct with us the end user at the top utilising Various Dsp's cpu's(inc arch variance) GPu's in every day ways easily and faster ,mantle is undoubtabely tied into Hsa due to Amd leaning toward APu's and an Amd Apu runs code optimised for an amd cpu better then intels so that initial parity (still enhanced for both) wont last.

 

[RoutedScripter]

Example of what Mantle would improve, in terms of CPU.

 

[Frag_Maniac]

Whilst initially id expect what your saying to hold out i dont think that will remain true with time ,who after all knows Amd's 64bit arch and instruction set best...

Except that AMD are smart enough thus far to have acknowledged that best case scenario is an abstraction layer as thin and universally seamless as possible, while still getting performance gains worthwhile enough to even pursue such a project. Otherwise the same thing will happen that happened with Glide. I think those overly enthusiastic about Mantle don't always realize the amount of caution Mantle's development was approached with, and I'm sure it has to do with the fact that is was requested by devs, vs initiated in house for AMD's exclusive benefit.

I'm sure the devs requesting it would rather the benefits of using it not be so drastically skewed to one side of their customer base. The common scenario and difference between games that sell well and those that don't is not how many are willing to buy it, but how many refuse to, and things like skewed performance can certainly cause the latter. There are ALWAYS going to be hardcore gamers whom have to have the latest titles, what differs is how many there are that cry foul and say not me.

 

[xenocide]

Example of what Mantle would improve, in terms of CPU.

Or Blizzard could make their engine use more than 1 thread, which would completely remove the CPU bottleneck.

 

[Frag_Maniac]

LOL, yeah single threading on games that are more data than graphics intense is really stupid anyway.

 

[RoutedScripter]

Except that AMD are smart enough thus far to have acknowledged that best case scenario is an abstraction layer as thin and universally seamless as possible, while still getting performance gains worthwhile enough to even pursue such a project. Otherwise the same thing will happen that happened with Glide. I think those overly enthusiastic about Mantle don't always realize the amount of caution Mantle's development was approached with, and I'm sure it has to do with the fact that is was requested by devs, vs initiated in house for AMD's exclusive benefit.

I'm sure the devs requesting it would rather the benefits of using it not be so drastically skewed to one side of their customer base. The common scenario and difference between games that sell well and those that don't is not how many are willing to buy it, but how many refuse to, and things like skewed performance can certainly cause the latter. There are ALWAYS going to be hardcore gamers whom have to have the latest titles, what differs is how many there are that cry foul and say not me.

I think you do not understand how much pressure has been made on the industry to get rid of API limitations on the PC. It is understandable, you have not researched and followed this as long as I was, there is still historical information available if you dig it. It is more of an insider-thing that doesn't creep out heavily in the mainstream gaming media, the naysayers are irrelevant noob-developers who show clear signs of enthusiastic-beginners, who use their novelty and instant-hit factor to influence the media which supercharges the exposure of these opinions they spew out.

While john carmack was a big part of these talks to get better development environment on PC, his priorities have been "compromised" (cannot criticize subjectively, he made no promises, his right to choose) by VR technology to the point his existing situation started to crumble apart, he is now considered totally irrelevant in this cause, he currently does not develop any game engine code, but I do not know if he has any contacts with the GPU vendors, if anything most probably VR-specific.



There will be no educated customer that would refuse to buy a mantle game purely based on opinion constructed with technical details. As an AMD engineer already said, Mantle's success is mostly limited by political and economical issues that may arise, for example, developers that wouldn't afford to be present in the transition period, or wouldn't believe it's a worthy enough thing, those mostly are irrelevant developer studios who have little influence on the shape of the industry, they are not required to be a part of the transition period.

Or Blizzard could make their engine use more than 1 thread, which would completely remove the CPU bottleneck.

You are wrong.

1. All DX applications have multi-threading utilization limitations.
2. It would not completely remove the bottleneck.
3. Starcraft 2 is not running on only 1 core, it does on 2, (but that's not real mutlithreading still, draw and physics probably run on one, and Ai on the other, overall it's not anywhere compared to some top FPS games.)

http://www.bit-tech.net/hardware/cpus/2010/08/18/how-many-cpu-cores-does-starcraft-2-use/1


First of all, the bottleneck always exists, it's just much higher in the performance graph, you will hit that bottleneck if you use the extra performance for more Ai, physics and other gameplay things.

Once games utilize this extra performance, the noticable boost gap will be lowered and eventually become nonexistant as the Ai, graphics and other things will consume this performance naturally, but this will take a few years.

The complexity of PC games will increase considerably in this transition to a new API, until another big increase comes, that might be quantum micro-chip technology.



EDIT:

 

[RoutedScripter]

Update:

Oops, that one slipped out.

http://forums.overclockers.co.uk/showpost.php?p=25445660&postcount=21

 

[TheoneandonlyMrK]

personally im not interested in a link to another thread on another forum with people bickering about mantle, I have that right here at home on TPU and since the name 8ball means nowt to me some context might have been helpfull

 

[RoutedScripter]

personally im not interested in a link to another thread on another forum with people bickering about mantle, I have that right here at home on TPU and since the name 8ball means nowt to me some context might have been helpfull

What are you saying, I haven't linked any thread to the stuff you mentioned, I linked to a post, and if you can't read who made that post than that's unfortunate for you. That post is a major indicator, now, those who don't understand mantle from the get go, obviously won't find this a big deal, this is news, but ofcourse, I don't really care what mainstream shit news decide to be "newsworthy".

After all this, I'm not really in a mood to bring up stuff for lazy people on a silver plate, I don't have any intention anymore to deeply discuss mantle with mantle haters/doubters/downplayers. I've exhausted what I meant in the 9 pages of this thread.

And who's 8ball ? It's 8 pack and if you checked a little more down you would see it's a staff member, and if you'd check the thread for info without paying attention to worthless posts than you'd know that someone from OCUK was at the AMD Summit and was rumored to be in the shown the closed demo from Oxide.

I have no trouble explaining, I have trouble because you lazy people expect it like I'm some kind of public service here.

 

[TheoneandonlyMrK]

What are you saying, I haven't linked any thread to the stuff you mentioned, I linked to a post, and if you can't read who made that post than that's unfortunate for you. That post is a major indicator, now, those who don't understand mantle from the get go, obviously won't find this a big deal, this is news, but ofcourse, I don't really care what mainstream shit news decide to be "newsworthy".

After all this, I'm not really in a mood to bring up stuff for lazy people on a silver plate, I don't have any intention anymore to deeply discuss mantle with mantle haters/doubters/downplayers. I've exhausted what I meant in the 9 pages of this thread.

And who's 8ball ? It's 8 pack and if you checked a little more down you would see it's a staff member, and if you'd check the thread for info without paying attention to worthless posts than you'd know that someone from OCUK was at the AMD Summit and was rumored to be in the shown the closed demo from Oxide.

I have no trouble explaining, I have trouble because you lazy people expect it like I'm some kind of public service here.

So my point was I don't know the guy and lazy I read a fair bit of that thread but your second post described it how you should have been wrote the first , why cryptic and why point to a vague dude all vague but not now thanks oh and im no mantle hater

 

[TheHunter]

OpenCL won't go anywhere, its here for totally different reasons - direct compute, meaning the gpu is used for normal computing, not just for graphics.

OpenCL provides parallel computing using task-based and data-based parallelism.

To some that still don't know, Mantle's only role is to have a thin layer between game and gpu, it can access all gpu features and much more beyond dx11.2 specifications.
Performance impact is minimal, it can have up to 100.000 directdraw calls @16ms which is 60fps, directx11 is capped at max ~ 12-16000 @ 16ms on average., openGL with bindless extensions up to 25000 calls. Consoles usually from 35000-50000 up to 100.000 or more.

So now you can have the same gfx detail and faster gameplay or richer worlds with more detail at steady 16ms - Im thinking CGI type of gfx.

Also it doesnt matter if a game uses lets say openCL physics or cuda physx DLL for physics simulations, it will coexist with this api just like direcX or openGL is now.



Anyway Mantle will rock you will see!

More devs need to step up, ie U4E, Luminous, Glacier2, Cryengine already is to some degree - it will be supported in Star Citizen, etc..

 

[Frag_Maniac]

I think you do not understand how much pressure has been made on the industry to get rid of API limitations on the PC. It is understandable, you have not researched and followed this as long as I was, there is still historical information available if you dig it. It is more of an insider-thing that doesn't creep out heavily in the mainstream gaming media, the naysayers are irrelevant noob-developers who show clear signs of enthusiastic-beginners, who use their novelty and instant-hit factor to influence the media which supercharges the exposure of these opinions they spew out.

While john carmack was a big part of these talks to get better development environment on PC, his priorities have been "compromised" (cannot criticize subjectively, he made no promises, his right to choose) by VR technology to the point his existing situation started to crumble apart, he is now considered totally irrelevant in this cause, he currently does not develop any game engine code, but I do not know if he has any contacts with the GPU vendors, if anything most probably VR-specific.


There will be no educated customer that would refuse to buy a mantle game purely based on opinion constructed with technical details. As an AMD engineer already said, Mantle's success is mostly limited by political and economical issues that may arise, for example, developers that wouldn't afford to be present in the transition period, or wouldn't believe it's a worthy enough thing, those mostly are irrelevant developer studios who have little influence on the shape of the industry, they are not required to be a part of the transition period.

I find it odd after all the people that have pointed out things you're not aware of or understanding of in this thread you still insist we're clueless. No one knows the exact numbers percentage wise of devs in the industry that have made a specific request for AMD or anyone else to come up with an alternative to Direct3D, and I'm sure most of them, unlike you, are understanding of the fact that if not implemented carefully, the same thing could happen that did with Glide. You keep glossing over that. Like I said earlier, you're ignorant of history in the industry in this regard. If you can't even grasp that simple concept, then no, you have not in fact "researched" this so well. I find it odd you'd even use the word research as naive as you sound.

If you had a shred of sense you'd angle this thread toward things that AMD/DICE have done to ensure Mantle will not suffer the same fate as Glide, but instead you insist it's a non issue. There really isn't enough proof that they've done much to make devs think Nvidia and their customers will support this, and THAT'S what matters most, not your narrow minded perception. If all that mattered were the opinions of a few naive AMD fanboys there'd be nothing to worry about, but there's a whole lot more to it than that. You keep insisting we're pessimistic skeptics. No, we're just realists that aren't easily convinced by a few stage presentations.